Polytab Analytics Blog

How to do multi-touch attribution?

You have asked the questions about multi-touch attribution in multiple 1-1 calls so this merited a blog post.

  • Do my affiliates deserve the commission?
  • What % of the revenue is generated by new vs existing customers?
  • How different are the purchase paths for those coming in through branded vs non-branded search?
  • What is the LTV (Long-term value) of my customers by the source of acquisition?

The answer to all of the questions can be answered by assessing the shopping journey leading up to the purchase for every order (across all channels).

Measuring multi-touch shopping impact

I will illustrate with the (sanitized) sample below downloaded from a customer’s Polytab attribution deployment. There are two orders whose sessions are listed in the screengrab.

The impact of a channel towards the purchase is the influence on the shopper’s purchase decision in respect to

  1. subsequent click intensity
  2. proximity to purchase
  3. raising awareness of the brand
  4. increasing interest
  5. impact on subsequent “direct” hits


Reference the two orders in the screen grab above.

For the order associated with shopping device 0x566089B4972D4C84478B8C3E

  1. The “bronto-email” gets some credit for the purchase 2.9% because it created awareness but there was no activity from December 3 to January 21.
  2. The “bing-organic” gets more credit 97% because it raised enough interest to merit a second hit at 10:10pm leading up to the purchase at 10:44pm.
  3. The attribution for the “direct” hit is redistributed to the previous touches

For the order associated with shopping device 0x56608AF5972D4C05488B989F

  1. The first touch is from “googleads-cpc” campaign (masked for confidentiality purpose) but it’s dated December 3. Based on the purchase on January 7, relatively less weight is given to it at 1.4%
  2. The second touch is also from “googleads-cpc”, albeit a different campaign (masked) and this gets a higher attribution at 50.47% due to proximity to the purchase and the subsequent clicks
  3. The next touch is from a branded “googleads-cpc” campaign and this was given a weight of 48%.
  4. Subsequent to the third touch the user was coming directly to the website and did over fifteen sessions to complete a purchase by 9:10pm on the same day.

In summary – multi-touch attribution is a bottom-up process

The point I am trying to make is that the weight of a touch to a campaign cannot be determined at an aggregate order level. The calculations have to be done at the order level and then aggregated up to the channel, campaign level. An even easier use-case is measuring if display brings new customers to your customer mix. If the display impression appears first in the touch/impression sequence, the shopper is a new customer, otherwise its a remarketed audience.

The multi-touch shopping journey

Want to discuss multi-touch attribution or see how you can view the complete multi-touch shopping journey on-demand? Contact me via the link below.

Sticky fingers image by Mark Freeth used with permission under creative commons license.